Cascading ecological effects of anthropogenic habitat fragmentation have been studied primarily in extreme cases (e.g., the isolation of habitat fragments in a novel habitat matrix such as suburban developments, reservoirs, or agricultural fields), with less attention to more subtle and widespread cases, such as habitat fragmentation due to timber harvest. Few studies have used rigorous demographic data to demonstrate the direct and indirect effects of habitat fragmentation. We trapped deer mice (Peromyscus maniculatus) at five sites over two years in southwest Oregon, USA, and used multi-state capture–recapture models to estimate deer mouse survival and movement in clearcuts, forest-fragment edges, forest-fragment interiors, and contiguous forests. We also estimated deer mouse densities in fragmented and unfragmented forests and combined deer mouse demographic studies with trillium (Trillium ovatum) seed predation trials to link deer mouse changes to reduced trillium recruitment previously observed at the same study sites. Mouse survival was highest in clearcuts, intermediate in forest fragments, and lowest in unfragmented (control) forests. Mouse movement among clearcuts, forest edges, and forest interiors was common over short time intervals. Collectively, demographic rates led to mouse densities that were 3–4 times higher at forest-fragment sites than at unfragmented sites. Trillium seeds were ∼3 times more likely to be depredated in areas of elevated relative mouse abundance than in areas of lower relative abundance. Forest fragmentation has favored mouse populations, resulting in increased seed predation that may decrease recruitment rates and increase local extinction risks for trillium.